从 Kubernetes 退出的 Pod 导致 Pod memory 使用量激增

Question

We have this python image running a Sanic server with a simple entrypoint:

ENTRYPOINT ["python3.9", "entrypoint.py"]

All of our orchestration is managed by Kubernetes.

Whenever a pod is deleted, the pod exits with a spike in memory usage, alerting our Grafana dashboards

How can I debug this?

Answer 1

its probably because the python process is doing something on exit thats eating ram. As a work around you can change grafana to a average ram usage across a time range to smooth out that spike